How can I benchmark C code easily?

Use the function clock() defined in time.h:

startTime = (float)clock()/CLOCKS_PER_SEC;/* Do work */endTime = (float)clock()/CLOCKS_PER_SEC;timeElapsed = endTime - startTime;

Basically, all you want is a high resolution timer. The elapsed time is of course just a difference in times and the speedup is calculated by dividing the times for each task. I have included the code for a high resolution timer that should work on at least windows and unix.

#ifdef WIN32#include <windows.h>double get_time(){    LARGE_INTEGER t, f;    QueryPerformanceCounter(&t);    QueryPerformanceFrequency(&f);    return (double)t.QuadPart/(double)f.QuadPart;}#else#include <sys/time.h>#include <sys/resource.h>double get_time(){    struct timeval t;    struct timezone tzp;    gettimeofday(&t, &tzp);    return t.tv_sec + t.tv_usec*1e-6;}#endif

Benchmark C code easily

#include <time.h>int main(void) {  clock_t start_time = clock();  // code or function to benchmark  double elapsed_time = (double)(clock() - start_time) / CLOCKS_PER_SEC;  printf("Done in %f seconds\n", elapsed_time);}

Easy benchmark of multi-threaded C code

If you want to benchmark multithreaded program you first need to take a closer look at clock:

Description

The clock() function returns an approximation of processor timeused by the program.

Return value

The value returned is the CPU time used so far as a clock_t; toget the number of seconds used, divide by CLOCKS_PER_SEC. If theprocessor time used is not available or its value cannot berepresented, the function returns the value (clock_t)(-1)

Hence it is very important to divide your elapsed_time by the number of threads in order to get the execution time of your function:

#include <time.h>#include <omp.h>#define THREADS_NB omp_get_max_threads()#pragma omp parallel for private(i) num_threads(THREADS_NB)clock_t start_time = clock();// code or function to benchmarkdouble elapsed_time = (double)(clock() - start_time) / CLOCKS_PER_SEC;printf("Done in %f seconds\n", elapsed_time / THREADS_NB); // divide by THREADS_NB!

Example

#include <stdlib.h>#include <string.h>#include <stdio.h>#include <time.h>#include <omp.h>#define N 20000#define THREADS_NB omp_get_max_threads()void init_arrays(double *a, double *b) {  memset(a, 0, sizeof(a));  memset(b, 0, sizeof(b));  for (int i = 0; i < N; i++) {    a[i] += 1.0;    b[i] += 1.0;  }}double func2(double i, double j) {  double res = 0.0;  while (i / j > 0.0) {    res += i / j;    i -= 0.1;    j -= 0.000003;  }  return res;}double single_thread(double *a, double *b) {  double res = 0;  int i, j;  for (i = 0; i < N; i++) {    for (j = 0; j < N; j++) {      if (i == j) continue;      res += func2(a[i], b[j]);    }  }  return res;}double multi_threads(double *a, double *b) {  double res = 0;  int i, j;  #pragma omp parallel for private(j) num_threads(THREADS_NB) reduction(+:res)  for (i = 0; i < N; i++) {    for (j = 0; j < N; j++) {      if (i == j) continue;      res += func2(a[i], b[j]);    }  }  return res;}int main(void) {  double *a, *b;  a = (double *)calloc(N, sizeof(double));  b = (double *)calloc(N, sizeof(double));  init_arrays(a, b);  clock_t start_time = clock();  double res = single_thread(a, b);  double elapsed_time = (double)(clock() - start_time) / CLOCKS_PER_SEC;  printf("Default:  Done with %f in %f sd\n", res, elapsed_time);  start_time = clock();  res = multi_threads(a, b);  elapsed_time = (double)(clock() - start_time) / CLOCKS_PER_SEC;  printf("With OMP: Done with %f in %f sd\n", res, elapsed_time / THREADS_NB);}

Compile with:

gcc -O3 multithread_benchmark.c -fopenmp && time ./a.out

Output:

Default:  Done with 2199909813.614555 in 4.909633 sdWith OMP: Done with 2199909799.377532 in 1.708831 sdreal    0m6.703s (from time function)